- Maya Barmecha
You have probably heard of ChatGTP (Generative Pre-trained Transformer). If you haven’t, ChatGTP is essentially a chatbot that uses generative artificial intelligence (AI) to answer all sorts of questions in accessible language. It was released by OpenAI on November 30th, 2022 and now has over 100 million users. If you read the previous issue of Campus, you may recall that ChatGTP wrote an article about itself for the issue. People all over the world are experimenting with ChatGTP to create emails, reviews, reports, and even produce code.
Well, less than four months into ChatGTP’s debut, Google recently released its take on an AI chatbot called Google Bard. Some say it may be a rival to OpenAI’s ChatGTP. So, we know Google Bard is a chatbot, but what makes it so special?
Google Bard AI is a natural language processing model developed by Google's research team. Bard is a state-of-the-art language model that uses machine learning algorithms to generate human-like text. A "bard" refers to a poet or storyteller. The model is designed to generate creative text.
Similar to ChatGTP, GTP-3 is a past language model that predicts future actions based on past information, making it autoregressive in computer parlance. OpenAI released it in 2020 and is the third generation of its model. Unlike ChatGTP, it is not easy to access and is mainly used for app development.
Google Bard’s purpose will be to provide more in-depth responses to searched questions than a classic Google search. Google Bard uses Google’s Language Model for Dialogue Applications (LaMDA), which is yet another chatbot, released by Google in May of 2021. Its function, like most chatbots, is to generate responses that are as close to human-generated as “mechanically” possible.
The significant difference between LaMDA and Bard would be that Bard uses less computing power, so it will be able to scale for more people to use it and provide feedback. In addition, Bard will work as a personal assistant and help with tasks like booking travel accommodations, searching for existing reservations, and helping with meal planning.
Google Bard’s demo was released on February 6th, where users found errors in the information outputted by the chatbot. A gif captured the application’s response to a question about a telescope in which Bard proceeded to display incorrect information about the telescope capturing “the very first pictures of a planet outside of our own solar system.”
Astrophysicists and astronomers responded to the gif, saying the chatbot was incorrect. The first pictures were taken years before the telescope was invented.
Some believe Google, to compete with ChatGTP, is rushing too much. Even Google employees believe that the launch was “rushed” and “botched” according to multiple news outlets.
Even so, as with all AI models, there are concerns about the potential misuse of such technology, such as creating fake news or propaganda. It is important to approach these models with caution and critical thinking. In addition, this is only the demo trial of the chatbot. We can anticipate a more polished official product.
ChatGTP and Google Bard share a lot in common with respect to their functionality. Similar to other chatbots, their main purpose is to answer users' questions in a style reminiscent of a human conversation.
The main difference lies in the data source of each chatbot. Bard replenishes its information from the internet, while ChatGTP only contains information from 2021 and earlier, so its range is limited. Bard also uses LaMDA for dialogue applications, while ChatGTP uses GTP 3.5. Bard will create more pieces of information, whereas ChatGTP creates content from a single text prompt.
Aside from comparisons and first impressions of Google Bard, only time will tell if Google Bard is a true competitor to chatbots such as ChatGTP.
NOTE: As I write this article, Meta’s new large language model, LLaMa has joined Microsoft’s ChatGPT. The newly-devised Sydney has entered the competition for the public’s favor in the current AI chatbox race.